EE/CNS 148 - Spring 2005

Lecture 1 (March 29)

Introduction to visual recognition. Recognition tasks: verification, detection and localization, classification.
Objects vs categories. Models composed of parts and geometry. Generative vs discriminative models. Learning: supervised, weakly supervised, unsupervised. Features: detection and description. Modeling
vs learning.

  1. Lecture

Lecture 2 (March 31)

Features and classifiers. Two types of errors. Conditional probability of each error. ROC curve. Linear classifiers.
The matched filter. Generalization to make it translation-invariant. Brief discussion of generalization to rotation, scale, lighting invariance.

  1. Lecture
  2. Matlab code for lecture: (a), (b), (c)

Lecture 3 (April 5)

Quadratic decision boundaries. Lambertian Models. Singular Value Decomposition. Principal components of face images.

  1. PCA on faces ppr by Turk and Pentland (Eigenfaces)
  2. lecture notes
  3. code from lecture

Lecture 4 (April 7)

Fisher Linear Discriminants. Comparison of PCA, FLD on face data-sets from Caltech.

  1. Lecture Notes
  2. Code for lecture 4 (Fisher Demo).
  3. Code for lecture 4 (Male Female Demo).

Lecture 5 (April 12)

Presented David Lowe Object Recognition system. Topics covered: key-points (aka interest points), SIFT features, KD-tree.

  1. David Lowe Paper (older)
  2. IJCV (new version, read this first).
  3. Lecture

Lecture 6 (April 14)

Review of Lowe Algorithm. Basic steps for training Lowe. Basic steps for testing. Hough Transform. Introduction to constellation model. Bayes Rule. Analogy to taking 'snapshot and detecting ships on an isolated island'.

  1. Joseph Gonzalez Eye Detector from HW 1.
  2. Lecture (Constellation Model)

Lecture 7 (April 19)

Continuation of constellation model. We presented simulations on a model composed of 3/4 parts. The optimal hypotheses were shown in a graphical manner. There was a full derivation of each term within the probabilistic framework for the constellation model.

  1. Code from Lecture

Lecture 8 (April 21)

Guest Lecture. Pierre Moreels (Perona Lab Graduate Student). Presents model which a cross between Constellation Model and Lowe Model. Recognizing Features in 3D.

  1. Presentation

Lecture 9 (Tues, April 26)

Guest lecture: Hsuan-Tien Lin ( from the Learning Systems Group at Caltech. Boosting. Joint Boosting. Adaptive Boosting. Paper on using Joint Boosting for Sharing Features.

  1. Presentation

Lecture 10 (Thurs, April 28)

Finished Boosting lecture. Ruxandra Paun presents Viola and Jones (both Facial Detection and Pedestrian Detection results).

  1. Presentation by Ruxandra.
  2. Movie of Viola and Jones Pedestrian detector.

Lecture 11 (Tues, May 3)

Marco Andreetto ( presents Liebe and Schiele ppr on object recognition.

Lecture 12 (Thurs, May 5)

Lecture 13 (Tues, May 10)

Lecture 14 (Thurs, May 12)

Lecture 15 (Tues, May 17)


Lecture 16 (Thurs, May 19)

Lecture 17 (Tues, May 24)

Lecture 18 (Thurs, May 26)

Final Project Review -- June 1, 11am-1pm in the Vision Lab.

Final Project Write-up due June 5th.